Simulated Operand Dataset

该数据集包含了用于训练强化学习代理解决算术问题的样本操作数，特别是涉及多数字的加法和减法，且这些操作数的范围超出了训练区间。此外，该代理在不同操作数区间的表现进行了评估，重点在于其分布外（OOD）的泛化能力。每个区间的样本量为10万，任务为解决算术问题。

This dataset contains sample operands for training reinforcement learning (RL) agents to solve arithmetic problems, specifically focusing on multi-digit addition and subtraction tasks where the operand ranges fall outside the training intervals. Additionally, we evaluate the agents' performance across different operand ranges, with particular emphasis on their out-of-distribution (OOD) generalization ability. Each range includes 100,000 samples, and the core task for the agents is to solve arithmetic problems.